最佳文本转语音模型

Gemini 3.8 Flash TTS is a text-to-speech model from Google and the successor to Gemini 3.1 Flash TTS Preview. It is the creative tier of the 3.8 TTS family, suited for narration, character work, and multi-speaker dialogue where voice fidelity, acting nuance, and regional dialect coverage matter more than latency.

It accepts the same speech configuration as its predecessor, including the 30 prebuilt studio voices, and supports Google's Voice Design (custom voices from a text description) and Voice Replication (voices cloned from a short sample) through voice_... identifiers.

Gemini 3.8 Flash Lite TTS is a text-to-speech model from Google and the fast, high-throughput member of the 3.8 TTS family alongside Gemini 3.8 Flash TTS. It is suited for conversational voice-agent pipelines, high-volume read-aloud, and voice replication workloads where latency and cost matter more than maximum expressiveness.

It accepts the same speech configuration as Gemini 3.1 Flash TTS Preview, including the 30 prebuilt studio voices, and supports Google's Voice Design and Voice Replication through voice_... identifiers.

Gemini 3.1 Flash TTS Preview is a text-to-speech model from Google, and a substantial generational step up from Gemini 2.5 Flash TTS. It takes text input and produces audio output across 70+ languages — nearly 3× the language coverage of its predecessor.

The headline addition is a system of 200+ inline audio tags (e.g. [whispers], [laughs], [excited]) that let developers steer delivery, emotion, and pacing mid-sentence, alongside a "director's chair" workflow in Google AI Studio for defining per-character Audio Profiles and scene-level context. It supports up to two speakers with independent voice and style configuration per speaker, outputs PCM audio at 24 kHz / 16-bit mono, and automatically watermarks all output with SynthID. Context window is 32k tokens.

Drama 3 is a text-to-speech model from Fish Audio built for directed voice performances. You write the direction in plain language in square brackets before a line, such as [nervous, barely holding it together], and can add sounds and pauses inline or change the delivery of specific words with paired tags like <whisper>.

It supports multi-speaker scenes in one request, switching language mid-sentence across 80+ languages in the same voice, and phoneme tags for pinning pronunciation of names and terms.

Kokoro 82M is a lightweight, open-weight text-to-speech model from hexgrad. It converts text to speech across 8 languages (American and British English, Spanish, French, Hindi, Italian, Japanese, Portuguese, and Chinese) using 54 preset voices organized by language and gender. At 82M parameters, it is well-suited for multilingual TTS deployments where footprint and cost efficiency matter.

Eleven v4 Turbo is a low-latency text-to-speech model from ElevenLabs. It keeps Eleven v4's expressive delivery and audio tags while being tuned for faster generation, with support for 90+ languages and a 10,000-character request limit.

Grok Voice TTS 1.0 is a text-to-speech model from SpaceXAI. It converts text into spoken audio across 20+ languages with automatic language detection, and offers five built-in voices (Eve, Ara, Rex, Sal, Leo) covering a range of tones. Inline speech tags allow control over pauses, emphasis, pitch, speed, and vocal style. Output is available in MP3, WAV, PCM, μ-law, and A-law formats at sample rates from 8 kHz to 48 kHz, with up to 15,000 characters per request.

Eleven v4 is a text-to-speech model from ElevenLabs. It is ElevenLabs' most expressive model, with inline audio tags for emotional and delivery control, support for 90+ languages, and a 10,000-character request limit.

HeyGen Voice Turbo is a lower-latency version of HeyGen Voice, HeyGen's text-to-speech model. It has the same 10 default voices, or bring your own HeyGen key to use voice clones from your own account. It streams 44.1 kHz mono PCM audio, with support for 39 languages and up to 5,000 characters per request.

S2.1 Pro Free is the no-cost variant of Fish Audio S2.1 Pro, intended for testing, prototyping, and low-volume applications. It provides the same synthesis capabilities without production latency or availability guarantees.

S2.1 Pro is a production-oriented text-to-speech model from Fish Audio. It is suited for multilingual voice applications, expressive narration, and dialogue synthesis, with open-ended natural-language controls for speaking style and emotion.

HeyGen Voice Turbo is a lower-latency version of HeyGen Voice, HeyGen's text-to-speech model. It has the same 10 default voices, or bring your own HeyGen key to use voice clones from your own account. It streams 44.1 kHz mono PCM audio, with support for 39 languages and up to 5,000 characters per request.

MAI-Voice-2 is an expressive text-to-speech model from Microsoft AI. It is suited for conversational assistants, media narration, accessibility, education, and other long-form voice applications. It supports 15 languages across 18 locales, fine-grained control of tone and delivery, multi-speaker generation, and voice prompting from short audio clips without fine-tuning. The model prioritizes naturalness and expressivity over latency-critical generation.

Eleven v3 is a text-to-speech model from ElevenLabs. It produces emotionally rich, highly expressive speech with inline audio tags, supports 70+ languages, and has a 5,000-character request limit.

Eleven Flash v2.5 is an ultra-low-latency text-to-speech model from ElevenLabs. It is suited for conversational and real-time use cases, supports 32 languages, and has a 40,000-character request limit.

Voxtral TTS 3 is Mistral's text-to-speech model. It converts text into natural-sounding speech, offers a library of preset voices, supports zero-shot voice cloning, and accepts emotion and pace tags in the input to control delivery. It streams audio with low time-to-first-audio and suits voice agents, narration, and multilingual assistants.

MAI-Voice-2.1-Flash is a low-latency text-to-speech model from Microsoft AI, optimized for real-time responsiveness. It produces natural, expressive speech across 23 languages, with human-like intonation, rhythm, and emotional nuance. It is suited for voice agents, assistants, call centers, and other interactive applications where latency and cost matter most.

On OpenRouter, set voice to a full voice ID with the model suffix, such as "en-US-Harper:MAI-Voice-2.1-Flash". A voice's locale sets the synthesis language. Set response_format to "mp3" or "pcm" (24 kHz mono). Harper and Grant support the agent, customer-call-center, educational, and narrator speaking styles, and many locale voices add emotion styles such as excited, happy, sad, and whispering. The full list of voices is in the supported_voices field of the models API. See the text-to-speech guide.

MAI-Voice-2.1 is Microsoft AI's highest-fidelity, most expressive text-to-speech model. It produces natural, studio-grade speech across 23 languages, with detailed prosody, nuanced expressiveness, and speaker consistency over long-form content. It is suited for audiobooks, podcasts, lectures, narration, and brand audio where maximum voice quality matters. The model prioritizes naturalness and expressivity over latency-critical generation.

On OpenRouter, set voice to a full voice ID with the model suffix, such as "en-US-Harper:MAI-Voice-2.1". A voice's locale sets the synthesis language. Set response_format to "mp3" or "pcm" (24 kHz mono). Harper and Grant support the agent, customer-call-center, educational, and narrator speaking styles, and many locale voices add emotion styles such as excited, happy, sad, and whispering. The full list of voices is in the supported_voices field of the models API. See the text-to-speech guide.

Eleven Multilingual v2 is a text-to-speech model from ElevenLabs. It is suited for lifelike, consistent long-form narration such as voice-overs and audiobooks, supports 29 languages, and has a 10,000-character request limit.

Qwen-Audio-3.0-TTS Plus is Alibaba's higher-quality text-to-speech model, generating spoken audio from text via the DashScope Speech Synthesizer API.

Seed Audio 1.0 is ByteDance Seed's non-streaming audio generation model. It produces speech and other audio from a natural-language text prompt that can describe the desired voice, tone, and sound effects, optionally guided by a Seed speaker ID or a reference audio clip for voice cloning. Output is limited to 120 seconds per request and is billed per second of generated audio. Suited to audiobooks, voiceovers, games, and similar workloads.

Qwen-Audio-3.0-TTS Flash is Alibaba's fast, cost-efficient text-to-speech model, generating spoken audio from text via the DashScope Speech Synthesizer API.

Eleven v3 Conversational is a text-to-speech model from ElevenLabs, a variant of Eleven v3 optimized for natural dialogue in conversational agents. It supports 70+ languages and has a 5,000-character request limit.

Aura-2 is a multilingual text-to-speech model from Deepgram. It supports Deepgram’s canonical Aura-2 voice catalog for speech synthesis across multiple languages.

Voxtral Mini TTS is Mistral's text-to-speech model featuring zero-shot voice cloning and multilingual support. It converts text input into natural-sounding audio output.